HITInfrastructure

Networking News

Artificial Intelligence Uses EHRs as Smart Analytics Tools

CloudMedX CEO, Tashfeen Suleman explains the value of natural language processing and machine learning for EHR analytics.

Source: Thinkstock

- As artificial intelligence (AI) continues to grow in health IT infrastructure, vendors are finding ways to develop AI solutions to improve patient care using the data collected by connected medical devices.

CloudMedX CEO Tashfeen Suleman developed his AI solution to improve clinician workflows by turning electronic health records (EHRs) into smart predictive tools, making doctors more accurate in decision metrics.

Healthcare organizations are challenged by the limitations of EHRs and the inability to truly leverage EHR data between the structured and unstructured data entered by a clinician. The use of natural language processing (NLP) and machine learning (ML) are playing a major role in EHR healthcare analytics.

Suleman began CloudMedX to prevent important patient details from being overlooked or lost in the data.

“EHRs haven't been on the market for long,” Suleman told HITInfrastructure.com. “They started with the Affordable Care Act and the intent is that all healthcare organizations adopt EHRs. EHRs have a 70 percent adoption rate, meaning 30 percent are still on paper charts. The process of digitization has started but that process is no older than maybe seven to 10 years.”

READ MORE: Health Catalyst Launches Open-Source Machine Learning

“We started thinking, if we could somehow tap into an EHR and turn the EHR into a smart tool, a smart, predictive tool that is run by AI making a doctor smarter in his workflow, smarter in his decision metrics and then smarter in treating and talking to his patients,” Suleman continued.

CloudMedX uses NLP to read through clinical notes in real time to give the notes context.

“For example, a patient says he has the flu and his father has a history of myocardial infarction and he is taking medication for diabetes,” Suleman explained. “The AI would look at that natural language processing and he'll know that the patient has the flu and that's a symptom. His father has a myocardial infarction, so he is at risk with a family history and he is taking a certain medication.”

Suleman noted the importance of NLP is to comprehend healthcare contextual analysis. The AI needs to know the context of words like cold, referring to an illness and not the weather. Beyond words with multiple meanings, the NLP also needs to understand the difference between chronic and present illnesses such as diabetes, to draw accurate conclusions from the data.

The NLP can analyze a patient’s entire medical history in real time and connect symptoms with current chronic illnesses, or an illness that runs in the patient’s family, potentially catching and treating an illness before it becomes life threatening.

READ MORE: How Machine Learning Can Improve Healthcare Cybersecurity

NLP is also the main AI tool used to decode unstructured data.

“Unstructured data is stored as notes and those notes are typically the doctor’s notes, discharge summaries, intake assessments, diagnosis notes, et cetera and they're all unstructured. NLP reads through those unstructured notes and free text notes and understand the information within those notes,” Suleman stated.

ML is also used to advise clinicians and make recommendations. The clinicians accepting and rejecting the AI suggestions teaches the ML technology, improving it over time.

ML uses evidence-based knowledge as a basis for its algorithm. ML learns from the clinician’s behavior to make connections based on patient symptoms and demographics to apply to future cases with similar symptoms and demographics.

AI deployments are new to health IT infrastructure and most organizations do not have them. The organizations that currently have AI solutions are introducing them to their infrastructure one project at a time.

READ MORE: Healthcare Natural Language Processing Expects Steady Growth

Suleman explained that CloudMedX currently targets nine disease risk states including hypertension, stroke, diabetes, and pneumonia.

“We are starting with these disease states because these are pretty chronic conditions and the majority, somewhere around 85 to 90 percent, of the population has one or more of these diseases and they always show up as comorbidities. Our target right now is these chronic, intensive or emergency situations and how we can manage them,” said Suleman.

Suleman and his team are focused on chronic conditions and how AI can predict the onset of a chronic disease, and the progression rate of that disease. They aim to work with healthcare organizations to cut costs for both patients and providers by diagnosing patients early and accurately.

 “For example, once a patient has kidney disease, the slippery slope towards end-stage renal disease (ESRD) is a matter of months,” explained Suleman. “Once a patient gets into ESRD, costs go up, the needs to go on dialysis. If the patient has a heart condition, they may not be able to go on dialysis because there is a 50 percent risk of mortality. There are too many parameters that come into play.”

“A doctor guesstimates patient risk and goes with a gut feel,” he continued. “Most of the time they're right but sometimes they're wrong. Even a false negative can cost someone his life or cost someone a lot of money if there is an error or misdiagnosis.”

AI solutions like CloudMedX use massive amounts of data and legacy data that organizations have been unable to use. Legacy data can be sorted to provide a backlog of information for clinicians.

Cloud-based EHRs are easier to work with than on-premise EHR systems because of the flexibility cloud storage offers.

 “Most cloud-based EHR's are single tenants and they have all the records in one place. We do a historical data dump once and all the data is connected to the patients using identifiers. If it's a multi-cell system and they have an on-premise deployment, then it's a bit of a challenge but it's not impossible,” added Suleman.

The goal of CloudMedX and many healthcare organizations currently testing AI use cases is to make use of all the data they collect for predictive analytics.

AI is young in the healthcare technology industry, but Suleman believes the technology will catch on relatively quickly, citing that many technology startups are putting some element of AI into their solutions.  AI fits with EHR technology and can leverage it to give organizations successful predictive analytics solutions.

In the next several years, Suleman hopes to expand CloudMedX beyond the nine specialized healthcare fields it currently works in. He believes that there will be a high demand for healthcare AI over the next several years as organizations face the demand for better analytics for faster and more accurate diagnoses. 

{snp_hitin_newsletter_widget}